;***************************************************************************** ;* quant-a.asm: x86 quantization and level-run ;***************************************************************************** ;* Copyright (C) 2005-2022 x264 project ;* ;* Authors: Loren Merritt ;* Fiona Glaser ;* Christian Heine ;* Oskar Arvidsson ;* Henrik Gramner ;* ;* This program is free software; you can redistribute it and/or modify ;* it under the terms of the GNU General Public License as published by ;* the Free Software Foundation; either version 2 of the License, or ;* (at your option) any later version. ;* ;* This program is distributed in the hope that it will be useful, ;* but WITHOUT ANY WARRANTY; without even the implied warranty of ;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the ;* GNU General Public License for more details. ;* ;* You should have received a copy of the GNU General Public License ;* along with this program; if not, write to the Free Software ;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA. ;* ;* This program is also available under a commercial proprietary license. ;* For more information, contact us at licensing@x264.com. ;***************************************************************************** %include "x86inc.asm" %include "x86util.asm" SECTION_RODATA 64 %if HIGH_BIT_DEPTH decimate_shuf_avx512: dd 0, 4, 8,12, 1, 5, 9,13, 2, 6,10,14, 3, 7,11,15 %else dequant_shuf_avx512: dw 0, 2, 4, 6, 8,10,12,14,16,18,20,22,24,26,28,30 dw 32,34,36,38,40,42,44,46,48,50,52,54,56,58,60,62 %endif %macro DQM4 3 dw %1, %2, %1, %2, %2, %3, %2, %3 %endmacro %macro DQM8 6 dw %1, %4, %5, %4, %1, %4, %5, %4 dw %4, %2, %6, %2, %4, %2, %6, %2 dw %5, %6, %3, %6, %5, %6, %3, %6 dw %4, %2, %6, %2, %4, %2, %6, %2 %endmacro dequant8_scale: DQM8 20, 18, 32, 19, 25, 24 DQM8 22, 19, 35, 21, 28, 26 DQM8 26, 23, 42, 24, 33, 31 DQM8 28, 25, 45, 26, 35, 33 DQM8 32, 28, 51, 30, 40, 38 DQM8 36, 32, 58, 34, 46, 43 dequant4_scale: DQM4 10, 13, 16 DQM4 11, 14, 18 DQM4 13, 16, 20 DQM4 14, 18, 23 DQM4 16, 20, 25 DQM4 18, 23, 29 decimate_mask_table4: db 0,3,2,6,2,5,5,9,1,5,4,8,5,8,8,12,1,4,4,8,4,7,7,11,4,8,7,11,8,11,11,15,1,4 db 3,7,4,7,7,11,3,7,6,10,7,10,10,14,4,7,7,11,7,10,10,14,7,11,10,14,11,14,14 db 18,0,4,3,7,3,6,6,10,3,7,6,10,7,10,10,14,3,6,6,10,6,9,9,13,6,10,9,13,10,13 db 13,17,4,7,6,10,7,10,10,14,6,10,9,13,10,13,13,17,7,10,10,14,10,13,13,17,10 db 14,13,17,14,17,17,21,0,3,3,7,3,6,6,10,2,6,5,9,6,9,9,13,3,6,6,10,6,9,9,13 db 6,10,9,13,10,13,13,17,3,6,5,9,6,9,9,13,5,9,8,12,9,12,12,16,6,9,9,13,9,12 db 12,16,9,13,12,16,13,16,16,20,3,7,6,10,6,9,9,13,6,10,9,13,10,13,13,17,6,9 db 9,13,9,12,12,16,9,13,12,16,13,16,16,20,7,10,9,13,10,13,13,17,9,13,12,16 db 13,16,16,20,10,13,13,17,13,16,16,20,13,17,16,20,17,20,20,24 chroma_dc_dct_mask_mmx: dw 0, 0,-1,-1, 0, 0,-1,-1 chroma_dc_dmf_mask_mmx: dw 0, 0,-1,-1, 0,-1,-1, 0 chroma_dc_dct_mask: dw 1, 1,-1,-1, 1, 1,-1,-1 chroma_dc_dmf_mask: dw 1, 1,-1,-1, 1,-1,-1, 1 %if HIGH_BIT_DEPTH==0 dct_coef_shuffle: %macro DCT_COEF_SHUFFLE 8 %assign y x %rep 8 %rep 7 %rotate (~(y>>7))&1 %assign y y<<((~(y>>7))&1) %endrep db %1*2 %rotate 1 %assign y y<<1 %endrep %endmacro %assign x 0 %rep 256 DCT_COEF_SHUFFLE 7, 6, 5, 4, 3, 2, 1, 0 %assign x x+1 %endrep %endif SECTION .text cextern pb_1 cextern pw_1 cextern pw_2 cextern pw_256 cextern pd_1 cextern pb_01 cextern pd_1024 cextern deinterleave_shufd cextern popcnt_table %macro QUANT_DC_START 2 movd xm%1, r1m ; mf movd xm%2, r2m ; bias %if cpuflag(avx2) vpbroadcastdct m%1, xm%1 vpbroadcastdct m%2, xm%2 %elif HIGH_BIT_DEPTH SPLATD m%1, m%1 SPLATD m%2, m%2 %elif cpuflag(sse4) ; ssse3, but not faster on conroe mova m5, [pb_01] pshufb m%1, m5 pshufb m%2, m5 %else SPLATW m%1, m%1 SPLATW m%2, m%2 %endif %endmacro %macro QUANT_END 0 xor eax, eax %if cpuflag(sse4) ptest m5, m5 %else ; !sse4 %if ARCH_X86_64 %if mmsize == 16 packsswb m5, m5 %endif movq rcx, m5 test rcx, rcx %else %if mmsize == 16 pxor m4, m4 pcmpeqb m5, m4 pmovmskb ecx, m5 cmp ecx, (1< 0 mov t0d, 8*(%2-2*%3) %%loop: %1 [r0+(t0 )*SIZEOF_PIXEL], [r1+t0*2 ], [r1+t0*2+ 8*%3], [r0+(t0+ 4*%3)*SIZEOF_PIXEL] %1 [r0+(t0+8*%3)*SIZEOF_PIXEL], [r1+t0*2+16*%3], [r1+t0*2+24*%3], [r0+(t0+12*%3)*SIZEOF_PIXEL] sub t0d, 16*%3 jge %%loop RET %else %if mmsize < 32 %1 [r0+(8*%3)*SIZEOF_PIXEL], [r1+16*%3], [r1+24*%3], [r0+(12*%3)*SIZEOF_PIXEL] %endif %1 [r0+(0 )*SIZEOF_PIXEL], [r1+0 ], [r1+ 8*%3], [r0+( 4*%3)*SIZEOF_PIXEL] RET %endif %endmacro %macro DEQUANT16_FLAT 2-5 mova m0, %1 psllw m0, m4 %assign i %0-2 %rep %0-1 %if i mova m %+ i, [r0+%2] pmullw m %+ i, m0 %else pmullw m0, [r0+%2] %endif mova [r0+%2], m %+ i %assign i i-1 %rotate 1 %endrep %endmacro %if ARCH_X86_64 DECLARE_REG_TMP 6,3,2 %else DECLARE_REG_TMP 2,0,1 %endif %macro DEQUANT_START 2 movifnidn t2d, r2m imul t0d, t2d, 0x2b shr t0d, 8 ; i_qbits = i_qp / 6 lea t1d, [t0*5] sub t2d, t0d sub t2d, t1d ; i_mf = i_qp % 6 shl t2d, %1 %if ARCH_X86_64 add r1, t2 ; dequant_mf[i_mf] %else add r1, r1mp ; dequant_mf[i_mf] mov r0, r0mp ; dct %endif sub t0d, %2 jl .rshift32 ; negative qbits => rightshift %endmacro ;----------------------------------------------------------------------------- ; void dequant_4x4( dctcoef dct[4][4], int dequant_mf[6][4][4], int i_qp ) ;----------------------------------------------------------------------------- %macro DEQUANT 3 cglobal dequant_%1x%1, 0,3,6 .skip_prologue: DEQUANT_START %2+2, %2 .lshift: movd xm2, t0d DEQUANT_LOOP DEQUANT16_L, %1*%1/4, %3 .rshift32: neg t0d mova m3, [pd_1] movd xm2, t0d pslld m3, xm2 pxor m4, m4 psrld m3, 1 DEQUANT_LOOP DEQUANT32_R, %1*%1/4, %3 %if HIGH_BIT_DEPTH == 0 && (notcpuflag(avx) || mmsize == 32) cglobal dequant_%1x%1_flat16, 0,3 movifnidn t2d, r2m %if %1 == 8 cmp t2d, 12 jl dequant_%1x%1 %+ SUFFIX %+ .skip_prologue sub t2d, 12 %endif imul t0d, t2d, 0x2b shr t0d, 8 ; i_qbits = i_qp / 6 lea t1d, [t0*5] sub t2d, t0d sub t2d, t1d ; i_mf = i_qp % 6 shl t2d, %2 %if ARCH_X86_64 lea r1, [dequant%1_scale] add r1, t2 %else lea r1, [dequant%1_scale + t2] %endif movifnidn r0, r0mp movd xm4, t0d %if %1 == 4 %if mmsize == 8 DEQUANT16_FLAT [r1], 0, 16 DEQUANT16_FLAT [r1+8], 8, 24 %elif mmsize == 16 DEQUANT16_FLAT [r1], 0, 16 %else vbroadcasti128 m0, [r1] psllw m0, xm4 pmullw m0, [r0] mova [r0], m0 %endif %elif mmsize == 8 DEQUANT16_FLAT [r1], 0, 8, 64, 72 DEQUANT16_FLAT [r1+16], 16, 24, 48, 56 DEQUANT16_FLAT [r1+16], 80, 88, 112, 120 DEQUANT16_FLAT [r1+32], 32, 40, 96, 104 %elif mmsize == 16 DEQUANT16_FLAT [r1], 0, 64 DEQUANT16_FLAT [r1+16], 16, 48, 80, 112 DEQUANT16_FLAT [r1+32], 32, 96 %else mova m1, [r1+ 0] mova m2, [r1+32] psllw m1, xm4 psllw m2, xm4 pmullw m0, m1, [r0+ 0] pmullw m3, m2, [r0+32] pmullw m4, m1, [r0+64] pmullw m5, m2, [r0+96] mova [r0+ 0], m0 mova [r0+32], m3 mova [r0+64], m4 mova [r0+96], m5 %endif RET %endif ; !HIGH_BIT_DEPTH && !AVX %endmacro ; DEQUANT %if HIGH_BIT_DEPTH INIT_XMM sse2 DEQUANT 4, 4, 2 DEQUANT 8, 6, 2 INIT_XMM xop DEQUANT 4, 4, 2 DEQUANT 8, 6, 2 INIT_YMM avx2 DEQUANT 4, 4, 4 DEQUANT 8, 6, 4 %else %if ARCH_X86_64 == 0 INIT_MMX mmx DEQUANT 4, 4, 1 DEQUANT 8, 6, 1 %endif INIT_XMM sse2 DEQUANT 4, 4, 2 DEQUANT 8, 6, 2 INIT_XMM avx DEQUANT 4, 4, 2 DEQUANT 8, 6, 2 INIT_XMM xop DEQUANT 4, 4, 2 DEQUANT 8, 6, 2 INIT_YMM avx2 DEQUANT 4, 4, 4 DEQUANT 8, 6, 4 %endif %macro DEQUANT_START_AVX512 1-2 0 ; shift, flat %if %2 == 0 movifnidn t2d, r2m %endif imul t0d, t2d, 0x2b shr t0d, 8 ; i_qbits = i_qp / 6 lea t1d, [t0*5] sub t2d, t0d sub t2d, t1d ; i_mf = i_qp % 6 shl t2d, %1 %if %2 %if ARCH_X86_64 %define dmf r1+t2 lea r1, [dequant8_scale] %else %define dmf t2+dequant8_scale %endif %elif ARCH_X86_64 %define dmf r1+t2 %else %define dmf r1 add r1, r1mp ; dequant_mf[i_mf] %endif movifnidn r0, r0mp %endmacro INIT_ZMM avx512 cglobal dequant_4x4, 0,3 DEQUANT_START_AVX512 6 mova m0, [dmf] %if HIGH_BIT_DEPTH pmaddwd m0, [r0] %endif sub t0d, 4 jl .rshift %if HIGH_BIT_DEPTH vpbroadcastd m1, t0d vpsllvd m0, m1 mova [r0], m0 %else vpbroadcastw ym1, t0d vpmovsdw ym0, m0 pmullw ym0, [r0] vpsllvw ym0, ym1 mova [r0], ym0 %endif RET .rshift: %if HIGH_BIT_DEPTH == 0 pmovzxwd m1, [r0] pmaddwd m0, m1 %endif mov r1d, 1<<31 shrx r1d, r1d, t0d ; 1 << (-i_qbits-1) neg t0d vpbroadcastd m1, r1d vpbroadcastd m2, t0d paddd m0, m1 vpsravd m0, m2 %if HIGH_BIT_DEPTH mova [r0], m0 %else vpmovsdw [r0], m0 %endif RET cglobal dequant_8x8, 0,3 DEQUANT_START_AVX512 8 mova m0, [dmf+0*64] mova m1, [dmf+1*64] mova m2, [dmf+2*64] mova m3, [dmf+3*64] %if HIGH_BIT_DEPTH pmaddwd m0, [r0+0*64] pmaddwd m1, [r0+1*64] pmaddwd m2, [r0+2*64] pmaddwd m3, [r0+3*64] %else mova m6, [dequant_shuf_avx512] %endif sub t0d, 6 jl .rshift %if HIGH_BIT_DEPTH vpbroadcastd m4, t0d vpsllvd m0, m4 vpsllvd m1, m4 vpsllvd m2, m4 vpsllvd m3, m4 jmp .end .rshift: %else vpbroadcastw m4, t0d vpermt2w m0, m6, m1 vpermt2w m2, m6, m3 pmullw m0, [r0] pmullw m2, [r0+64] vpsllvw m0, m4 vpsllvw m2, m4 mova [r0], m0 mova [r0+64], m2 RET .rshift: pmovzxwd m4, [r0+0*32] pmovzxwd m5, [r0+1*32] pmaddwd m0, m4 pmaddwd m1, m5 pmovzxwd m4, [r0+2*32] pmovzxwd m5, [r0+3*32] pmaddwd m2, m4 pmaddwd m3, m5 %endif mov r1d, 1<<31 shrx r1d, r1d, t0d ; 1 << (-i_qbits-1) neg t0d vpbroadcastd m4, r1d vpbroadcastd m5, t0d paddd m0, m4 paddd m1, m4 vpsravd m0, m5 vpsravd m1, m5 paddd m2, m4 paddd m3, m4 vpsravd m2, m5 vpsravd m3, m5 %if HIGH_BIT_DEPTH .end: mova [r0+0*64], m0 mova [r0+1*64], m1 mova [r0+2*64], m2 mova [r0+3*64], m3 %else vpermt2w m0, m6, m1 vpermt2w m2, m6, m3 mova [r0], m0 mova [r0+64], m2 %endif RET %if HIGH_BIT_DEPTH == 0 cglobal dequant_8x8_flat16, 0,3 movifnidn t2d, r2m cmp t2d, 12 jl dequant_8x8_avx512 sub t2d, 12 DEQUANT_START_AVX512 6, 1 vpbroadcastw m0, t0d mova m1, [dmf] vpsllvw m1, m0 pmullw m0, m1, [r0] pmullw m1, [r0+64] mova [r0], m0 mova [r0+64], m1 RET %endif %undef dmf %macro DEQUANT_DC 2 cglobal dequant_4x4dc, 0,3,6 DEQUANT_START 6, 6 .lshift: %if cpuflag(avx2) vpbroadcastdct m3, [r1] %else movd xm3, [r1] SPLAT%1 m3, xm3 %endif movd xm2, t0d pslld m3, xm2 %assign %%x 0 %rep SIZEOF_PIXEL*32/mmsize %2 m0, m3, [r0+%%x] mova [r0+%%x], m0 %assign %%x %%x+mmsize %endrep RET .rshift32: neg t0d %if cpuflag(avx2) vpbroadcastdct m2, [r1] %else movd xm2, [r1] %endif mova m5, [p%1_1] movd xm3, t0d pslld m4, m5, xm3 psrld m4, 1 %if HIGH_BIT_DEPTH %if notcpuflag(avx2) pshufd m2, m2, 0 %endif %assign %%x 0 %rep SIZEOF_PIXEL*32/mmsize pmadcswd m0, m2, [r0+%%x], m4 psrad m0, xm3 mova [r0+%%x], m0 %assign %%x %%x+mmsize %endrep %else ; !HIGH_BIT_DEPTH %if notcpuflag(avx2) PSHUFLW m2, m2, 0 %endif punpcklwd m2, m4 %assign %%x 0 %rep SIZEOF_PIXEL*32/mmsize mova m0, [r0+%%x] punpckhwd m1, m0, m5 punpcklwd m0, m5 pmaddwd m0, m2 pmaddwd m1, m2 psrad m0, xm3 psrad m1, xm3 packssdw m0, m1 mova [r0+%%x], m0 %assign %%x %%x+mmsize %endrep %endif ; !HIGH_BIT_DEPTH RET %endmacro %if HIGH_BIT_DEPTH INIT_XMM sse2 DEQUANT_DC d, pmaddwd INIT_XMM xop DEQUANT_DC d, pmaddwd INIT_YMM avx2 DEQUANT_DC d, pmaddwd %else %if ARCH_X86_64 == 0 INIT_MMX mmx2 DEQUANT_DC w, pmullw %endif INIT_XMM sse2 DEQUANT_DC w, pmullw INIT_XMM avx DEQUANT_DC w, pmullw INIT_YMM avx2 DEQUANT_DC w, pmullw %endif %macro PEXTRW 4 %if cpuflag(sse4) pextrw %1, %2, %3 %else ; pextrw with a memory destination requires SSE4.1, go through a GPR as a fallback %if %3 pextrw %4d, %2, %3 %else movd %4d, %2 %endif mov %1, %4w %endif %endmacro ;----------------------------------------------------------------------------- ; void idct_dequant_2x4_dc( dctcoef dct[8], dctcoef dct4x4[8][16], int dequant_mf[6][16], int i_qp ) ; void idct_dequant_2x4_dconly( dctcoef dct[8], int dequant_mf[6][16], int i_qp ) ;----------------------------------------------------------------------------- %macro DEQUANT_2x4_DC 1 %ifidn %1, dconly DECLARE_REG_TMP 6,3,2 %define %%args dct, dmf, qp %else DECLARE_REG_TMP 6,4,3 %define %%args dct, dct4x4, dmf, qp %endif %if ARCH_X86_64 == 0 DECLARE_REG_TMP 2,0,1 %endif cglobal idct_dequant_2x4_%1, 0,3,5, %%args movifnidn t2d, qpm imul t0d, t2d, 0x2b shr t0d, 8 ; qp / 6 lea t1d, [t0*5] sub t2d, t0d sub t2d, t1d ; qp % 6 shl t2d, 6 ; 16 * sizeof(int) %if ARCH_X86_64 imul t2d, [dmfq+t2], -0xffff ; (-dmf) << 16 | dmf %else mov dctq, dctmp add t2, dmfmp imul t2d, [t2], -0xffff %endif %if HIGH_BIT_DEPTH mova m0, [dctq] mova m1, [dctq+16] SUMSUB_BA d, 1, 0, 2 ; 16-bit intermediate precision is enough for the first two sumsub steps, packssdw m1, m0 ; and by packing to words we can use pmaddwd instead of pmulld later. %else movq m0, [dctq] movq m1, [dctq+8] SUMSUB_BA w, 1, 0, 2 punpcklqdq m1, m0 ; a0 a1 a2 a3 a4 a5 a6 a7 %endif pshufd m0, m1, q2301 ; a2 a3 a0 a1 a6 a7 a4 a5 movd m3, t2d pshuflw m3, m3, q1000 ; + + + - SUMSUB_BA w, 0, 1, 2 punpcklqdq m3, m3 ; + + + - + + + - pshufd m1, m1, q0022 sub t0d, 6 jl .rshift movd m2, t0d psllw m3, m2 pmaddwd m0, m3 pmaddwd m1, m3 jmp .end .rshift: neg t0d movd m2, t0d pcmpeqd m4, m4 pmaddwd m0, m3 pmaddwd m1, m3 pslld m4, m2 psrad m4, 1 psubd m0, m4 ; + 1 << (qp/6-1) psubd m1, m4 psrad m0, m2 psrad m1, m2 .end: %ifidn %1, dconly %if HIGH_BIT_DEPTH mova [dctq], m0 mova [dctq+16], m1 %else packssdw m0, m1 mova [dctq], m0 %endif %else movifnidn dct4x4q, dct4x4mp %if HIGH_BIT_DEPTH movd [dct4x4q+0*64], m0 %if cpuflag(sse4) pextrd [dct4x4q+1*64], m0, 1 add dct4x4q, 4*64 pextrd [dct4x4q-2*64], m0, 2 pextrd [dct4x4q-1*64], m0, 3 movd [dct4x4q+0*64], m1 pextrd [dct4x4q+1*64], m1, 1 pextrd [dct4x4q+2*64], m1, 2 pextrd [dct4x4q+3*64], m1, 3 %else MOVHL m2, m0 psrlq m0, 32 movd [dct4x4q+1*64], m0 add dct4x4q, 4*64 movd [dct4x4q-2*64], m2 psrlq m2, 32 movd [dct4x4q-1*64], m2 movd [dct4x4q+0*64], m1 MOVHL m2, m1 psrlq m1, 32 movd [dct4x4q+1*64], m1 movd [dct4x4q+2*64], m2 psrlq m2, 32 movd [dct4x4q+3*64], m2 %endif %else PEXTRW [dct4x4q+0*32], m0, 0, eax PEXTRW [dct4x4q+1*32], m0, 2, eax PEXTRW [dct4x4q+2*32], m0, 4, eax PEXTRW [dct4x4q+3*32], m0, 6, eax add dct4x4q, 4*32 PEXTRW [dct4x4q+0*32], m1, 0, eax PEXTRW [dct4x4q+1*32], m1, 2, eax PEXTRW [dct4x4q+2*32], m1, 4, eax PEXTRW [dct4x4q+3*32], m1, 6, eax %endif %endif RET %endmacro ; sse4 reduces code size compared to sse2 but isn't any faster, so just go with sse2+avx INIT_XMM sse2 DEQUANT_2x4_DC dc DEQUANT_2x4_DC dconly INIT_XMM avx DEQUANT_2x4_DC dc DEQUANT_2x4_DC dconly ; t4 is eax for return value. %if ARCH_X86_64 DECLARE_REG_TMP 0,1,2,3,6,4 ; Identical for both Windows and *NIX %else DECLARE_REG_TMP 4,1,2,3,0,5 %endif ;----------------------------------------------------------------------------- ; x264_optimize_chroma_2x2_dc( dctcoef dct[4], int dequant_mf ) ;----------------------------------------------------------------------------- %macro OPTIMIZE_CHROMA_2x2_DC 0 cglobal optimize_chroma_2x2_dc, 0,6-cpuflag(sse4),7 movifnidn t0, r0mp movd m2, r1m movq m1, [t0] %if cpuflag(sse4) pcmpeqb m4, m4 pslld m4, 11 %else pxor m4, m4 %endif %if cpuflag(ssse3) mova m3, [chroma_dc_dct_mask] mova m5, [chroma_dc_dmf_mask] %else mova m3, [chroma_dc_dct_mask_mmx] mova m5, [chroma_dc_dmf_mask_mmx] %endif pshuflw m2, m2, 0 pshufd m0, m1, q0101 ; 1 0 3 2 1 0 3 2 punpcklqdq m2, m2 punpcklqdq m1, m1 ; 3 2 1 0 3 2 1 0 mova m6, [pd_1024] ; 32<<5, elements are shifted 5 bits to the left PSIGNW m0, m3 ; -1 -0 3 2 -1 -0 3 2 PSIGNW m2, m5 ; + - - + - - + + paddw m0, m1 ; -1+3 -0+2 1+3 0+2 -1+3 -0+2 1+3 0+2 pmaddwd m0, m2 ; 0-1-2+3 0-1+2-3 0+1-2-3 0+1+2+3 * dmf punpcklwd m1, m1 psrad m2, 16 ; + - - + mov t1d, 3 paddd m0, m6 xor t4d, t4d %if notcpuflag(ssse3) psrad m1, 31 ; has to be 0 or -1 in order for PSIGND_MMX to work correctly %endif %if cpuflag(sse4) ptest m0, m4 %else mova m6, m0 SWAP 0, 6 psrad m6, 11 pcmpeqd m6, m4 pmovmskb t5d, m6 cmp t5d, 0xffff %endif jz .ret ; if the DC coefficients already round to zero, terminate early mova m3, m0 .outer_loop: movsx t3d, word [t0+2*t1] ; dct[coeff] pshufd m6, m1, q3333 pshufd m1, m1, q2100 ; move the next element to high dword PSIGND m5, m2, m6 test t3d, t3d jz .loop_end .outer_loop_0: mov t2d, t3d sar t3d, 31 or t3d, 1 .inner_loop: psubd m3, m5 ; coeff -= sign pxor m6, m0, m3 %if cpuflag(sse4) ptest m6, m4 %else psrad m6, 11 pcmpeqd m6, m4 pmovmskb t5d, m6 cmp t5d, 0xffff %endif jz .round_coeff paddd m3, m5 ; coeff += sign mov t4d, 1 .loop_end: dec t1d jz .last_coeff pshufd m2, m2, q1320 ; - + - + / - - + + jg .outer_loop .ret: REP_RET .round_coeff: sub t2d, t3d mov [t0+2*t1], t2w jnz .inner_loop jmp .loop_end .last_coeff: movsx t3d, word [t0] punpcklqdq m2, m2 ; + + + + PSIGND m5, m2, m1 test t3d, t3d jnz .outer_loop_0 RET %endmacro %if HIGH_BIT_DEPTH == 0 INIT_XMM sse2 OPTIMIZE_CHROMA_2x2_DC INIT_XMM ssse3 OPTIMIZE_CHROMA_2x2_DC INIT_XMM sse4 OPTIMIZE_CHROMA_2x2_DC INIT_XMM avx OPTIMIZE_CHROMA_2x2_DC %endif ; !HIGH_BIT_DEPTH %if HIGH_BIT_DEPTH ;----------------------------------------------------------------------------- ; void denoise_dct( int32_t *dct, uint32_t *sum, uint32_t *offset, int size ) ;----------------------------------------------------------------------------- %macro DENOISE_DCT 0 cglobal denoise_dct, 4,4,6 pxor m5, m5 movsxdifnidn r3, r3d .loop: mova m2, [r0+r3*4-2*mmsize] mova m3, [r0+r3*4-1*mmsize] ABSD m0, m2 ABSD m1, m3 paddd m4, m0, [r1+r3*4-2*mmsize] psubd m0, [r2+r3*4-2*mmsize] mova [r1+r3*4-2*mmsize], m4 paddd m4, m1, [r1+r3*4-1*mmsize] psubd m1, [r2+r3*4-1*mmsize] mova [r1+r3*4-1*mmsize], m4 pcmpgtd m4, m0, m5 pand m0, m4 pcmpgtd m4, m1, m5 pand m1, m4 PSIGND m0, m2 PSIGND m1, m3 mova [r0+r3*4-2*mmsize], m0 mova [r0+r3*4-1*mmsize], m1 sub r3d, mmsize/2 jg .loop RET %endmacro %if ARCH_X86_64 == 0 INIT_MMX mmx DENOISE_DCT %endif INIT_XMM sse2 DENOISE_DCT INIT_XMM ssse3 DENOISE_DCT INIT_XMM avx DENOISE_DCT INIT_YMM avx2 DENOISE_DCT %else ; !HIGH_BIT_DEPTH ;----------------------------------------------------------------------------- ; void denoise_dct( int16_t *dct, uint32_t *sum, uint16_t *offset, int size ) ;----------------------------------------------------------------------------- %macro DENOISE_DCT 0 cglobal denoise_dct, 4,4,7 pxor m6, m6 movsxdifnidn r3, r3d .loop: mova m2, [r0+r3*2-2*mmsize] mova m3, [r0+r3*2-1*mmsize] ABSW m0, m2, sign ABSW m1, m3, sign psubusw m4, m0, [r2+r3*2-2*mmsize] psubusw m5, m1, [r2+r3*2-1*mmsize] PSIGNW m4, m2 PSIGNW m5, m3 mova [r0+r3*2-2*mmsize], m4 mova [r0+r3*2-1*mmsize], m5 punpcklwd m2, m0, m6 punpcklwd m3, m1, m6 punpckhwd m0, m6 punpckhwd m1, m6 paddd m2, [r1+r3*4-4*mmsize] paddd m0, [r1+r3*4-3*mmsize] paddd m3, [r1+r3*4-2*mmsize] paddd m1, [r1+r3*4-1*mmsize] mova [r1+r3*4-4*mmsize], m2 mova [r1+r3*4-3*mmsize], m0 mova [r1+r3*4-2*mmsize], m3 mova [r1+r3*4-1*mmsize], m1 sub r3, mmsize jg .loop RET %endmacro %if ARCH_X86_64 == 0 INIT_MMX mmx DENOISE_DCT %endif INIT_XMM sse2 DENOISE_DCT INIT_XMM ssse3 DENOISE_DCT INIT_XMM avx DENOISE_DCT INIT_YMM avx2 cglobal denoise_dct, 4,4,4 pxor m3, m3 movsxdifnidn r3, r3d .loop: mova m1, [r0+r3*2-mmsize] pabsw m0, m1 psubusw m2, m0, [r2+r3*2-mmsize] vpermq m0, m0, q3120 psignw m2, m1 mova [r0+r3*2-mmsize], m2 punpcklwd m1, m0, m3 punpckhwd m0, m3 paddd m1, [r1+r3*4-2*mmsize] paddd m0, [r1+r3*4-1*mmsize] mova [r1+r3*4-2*mmsize], m1 mova [r1+r3*4-1*mmsize], m0 sub r3, mmsize/2 jg .loop RET %endif ; !HIGH_BIT_DEPTH ;----------------------------------------------------------------------------- ; int decimate_score( dctcoef *dct ) ;----------------------------------------------------------------------------- %macro DECIMATE_MASK 4 %if HIGH_BIT_DEPTH mova m0, [%3+0*16] packssdw m0, [%3+1*16] mova m1, [%3+2*16] packssdw m1, [%3+3*16] ABSW2 m0, m1, m0, m1, m3, m4 %else ABSW m0, [%3+ 0], m3 ABSW m1, [%3+16], m4 %endif packsswb m0, m1 pxor m2, m2 pcmpeqb m2, m0 pcmpgtb m0, %4 pmovmskb %1, m2 pmovmskb %2, m0 %endmacro %macro DECIMATE_MASK16_AVX512 0 mova m0, [r0] %if HIGH_BIT_DEPTH vptestmd k0, m0, m0 pabsd m0, m0 vpcmpud k1, m0, [pd_1] {1to16}, 6 %else vptestmw k0, m0, m0 pabsw m0, m0 vpcmpuw k1, m0, [pw_1], 6 %endif %endmacro %macro SHRX 2 %if cpuflag(bmi2) shrx %1, %1, %2 %else shr %1, %2b ; %2 has to be rcx/ecx %endif %endmacro %macro BLSR 2 %if cpuflag(bmi1) blsr %1, %2 %else lea %1, [%2-1] and %1, %2 %endif %endmacro cextern_common decimate_table4 cextern_common decimate_table8 %macro DECIMATE4x4 1 cglobal decimate_score%1, 1,3 %if cpuflag(avx512) DECIMATE_MASK16_AVX512 xor eax, eax kmovw edx, k0 %if %1 == 15 shr edx, 1 %else test edx, edx %endif jz .ret ktestw k1, k1 jnz .ret9 %else DECIMATE_MASK edx, eax, r0, [pb_1] xor edx, 0xffff jz .ret test eax, eax jnz .ret9 %if %1 == 15 shr edx, 1 %endif %endif %if ARCH_X86_64 lea r4, [decimate_mask_table4] %define mask_table r4 %else %define mask_table decimate_mask_table4 %endif movzx ecx, dl movzx eax, byte [mask_table + rcx] %if ARCH_X86_64 xor edx, ecx jz .ret %if cpuflag(lzcnt) lzcnt ecx, ecx lea r5, [decimate_table4-32] add r5, rcx %else bsr ecx, ecx lea r5, [decimate_table4-1] sub r5, rcx %endif %define table r5 %else cmp edx, ecx jz .ret bsr ecx, ecx shr edx, 1 SHRX edx, ecx %define table decimate_table4 %endif tzcnt ecx, edx shr edx, 1 SHRX edx, ecx add al, byte [table + rcx] add al, byte [mask_table + rdx] .ret: REP_RET .ret9: mov eax, 9 RET %endmacro %macro DECIMATE_MASK64_AVX2 2 ; nz_low, nz_high mova m0, [r0+0*32] packsswb m0, [r0+1*32] mova m1, [r0+2*32] packsswb m1, [r0+3*32] mova m4, [pb_1] pabsb m2, m0 pabsb m3, m1 por m2, m3 ; the > 1 checks don't care about order, so ptest m4, m2 ; we can save latency by doing them here jnc .ret9 vpermq m0, m0, q3120 vpermq m1, m1, q3120 pxor m4, m4 pcmpeqb m0, m4 pcmpeqb m1, m4 pmovmskb %1, m0 pmovmskb %2, m1 %endmacro %macro DECIMATE_MASK64_AVX512 0 mova m0, [r0] %if HIGH_BIT_DEPTH packssdw m0, [r0+1*64] mova m1, [r0+2*64] packssdw m1, [r0+3*64] packsswb m0, m1 vbroadcasti32x4 m1, [pb_1] pabsb m2, m0 vpcmpub k0, m2, m1, 6 ktestq k0, k0 jnz .ret9 mova m1, [decimate_shuf_avx512] vpermd m0, m1, m0 vptestmb k1, m0, m0 %else mova m1, [r0+64] vbroadcasti32x4 m3, [pb_1] packsswb m2, m0, m1 pabsb m2, m2 vpcmpub k0, m2, m3, 6 ktestq k0, k0 jnz .ret9 vptestmw k1, m0, m0 vptestmw k2, m1, m1 %endif %endmacro %macro DECIMATE8x8 0 %if ARCH_X86_64 cglobal decimate_score64, 1,5 %if mmsize == 64 DECIMATE_MASK64_AVX512 xor eax, eax %if HIGH_BIT_DEPTH kmovq r1, k1 test r1, r1 jz .ret %else kortestd k1, k2 jz .ret kunpckdq k1, k2, k1 kmovq r1, k1 %endif %elif mmsize == 32 DECIMATE_MASK64_AVX2 r1d, eax not r1 shl rax, 32 xor r1, rax jz .ret %else mova m5, [pb_1] DECIMATE_MASK r1d, eax, r0+SIZEOF_DCTCOEF* 0, m5 test eax, eax jnz .ret9 DECIMATE_MASK r2d, eax, r0+SIZEOF_DCTCOEF*16, m5 shl r2d, 16 or r1d, r2d DECIMATE_MASK r2d, r3d, r0+SIZEOF_DCTCOEF*32, m5 shl r2, 32 or eax, r3d or r1, r2 DECIMATE_MASK r2d, r3d, r0+SIZEOF_DCTCOEF*48, m5 not r1 shl r2, 48 xor r1, r2 jz .ret add eax, r3d jnz .ret9 %endif lea r4, [decimate_table8] mov al, -6 .loop: tzcnt rcx, r1 add al, byte [r4 + rcx] jge .ret9 shr r1, 1 SHRX r1, rcx %if cpuflag(bmi2) test r1, r1 %endif jnz .loop add al, 6 .ret: REP_RET .ret9: mov eax, 9 RET %else ; ARCH cglobal decimate_score64, 1,4 %if mmsize == 64 DECIMATE_MASK64_AVX512 xor eax, eax %if HIGH_BIT_DEPTH kshiftrq k2, k1, 32 %endif kmovd r2, k1 kmovd r3, k2 test r2, r2 jz .tryret %elif mmsize == 32 DECIMATE_MASK64_AVX2 r2, r3 xor eax, eax not r3 xor r2, -1 jz .tryret %else mova m5, [pb_1] DECIMATE_MASK r2, r1, r0+SIZEOF_DCTCOEF* 0, m5 test r1, r1 jnz .ret9 DECIMATE_MASK r3, r1, r0+SIZEOF_DCTCOEF*16, m5 not r2 shl r3, 16 xor r2, r3 mov r0m, r2 DECIMATE_MASK r3, r2, r0+SIZEOF_DCTCOEF*32, m5 or r2, r1 DECIMATE_MASK r1, r0, r0+SIZEOF_DCTCOEF*48, m5 add r0, r2 jnz .ret9 mov r2, r0m not r3 shl r1, 16 xor r3, r1 test r2, r2 jz .tryret %endif mov al, -6 .loop: tzcnt ecx, r2 add al, byte [decimate_table8 + ecx] jge .ret9 sub ecx, 31 ; increase the shift count by one to shift away the lowest set bit as well jz .run31 ; only bits 0-4 are used so we have to explicitly handle the case of 1<<31 shrd r2, r3, cl SHRX r3, ecx %if notcpuflag(bmi2) test r2, r2 %endif jnz .loop BLSR r2, r3 jz .end .largerun: tzcnt ecx, r3 shr r3, 1 SHRX r3, ecx .loop2: tzcnt ecx, r3 add al, byte [decimate_table8 + ecx] jge .ret9 shr r3, 1 SHRX r3, ecx .run31: test r3, r3 jnz .loop2 .end: add al, 6 RET .tryret: BLSR r2, r3 jz .ret mov al, -6 jmp .largerun .ret9: mov eax, 9 .ret: REP_RET %endif ; ARCH %endmacro INIT_XMM sse2 DECIMATE4x4 15 DECIMATE4x4 16 DECIMATE8x8 INIT_XMM ssse3 DECIMATE4x4 15 DECIMATE4x4 16 DECIMATE8x8 %if HIGH_BIT_DEPTH INIT_ZMM avx512 %else INIT_YMM avx2 DECIMATE8x8 INIT_YMM avx512 %endif DECIMATE4x4 15 DECIMATE4x4 16 INIT_ZMM avx512 DECIMATE8x8 ;----------------------------------------------------------------------------- ; int coeff_last( dctcoef *dct ) ;----------------------------------------------------------------------------- %macro BSR 3 %if cpuflag(lzcnt) lzcnt %1, %2 xor %1, %3 %else bsr %1, %2 %endif %endmacro %macro LZCOUNT 3 %if cpuflag(lzcnt) lzcnt %1, %2 %else bsr %1, %2 xor %1, %3 %endif %endmacro %if HIGH_BIT_DEPTH %macro LAST_MASK 3-4 %if %1 == 4 movq mm0, [%3] packssdw mm0, [%3+8] packsswb mm0, mm0 pcmpeqb mm0, mm2 pmovmskb %2, mm0 %elif mmsize == 16 movdqa xmm0, [%3+ 0] %if %1 == 8 packssdw xmm0, [%3+16] packsswb xmm0, xmm0 %else movdqa xmm1, [%3+32] packssdw xmm0, [%3+16] packssdw xmm1, [%3+48] packsswb xmm0, xmm1 %endif pcmpeqb xmm0, xmm2 pmovmskb %2, xmm0 %elif %1 == 8 movq mm0, [%3+ 0] movq mm1, [%3+16] packssdw mm0, [%3+ 8] packssdw mm1, [%3+24] packsswb mm0, mm1 pcmpeqb mm0, mm2 pmovmskb %2, mm0 %else movq mm0, [%3+ 0] movq mm1, [%3+16] packssdw mm0, [%3+ 8] packssdw mm1, [%3+24] movq mm3, [%3+32] movq mm4, [%3+48] packssdw mm3, [%3+40] packssdw mm4, [%3+56] packsswb mm0, mm1 packsswb mm3, mm4 pcmpeqb mm0, mm2 pcmpeqb mm3, mm2 pmovmskb %2, mm0 pmovmskb %4, mm3 shl %4, 8 or %2, %4 %endif %endmacro %macro COEFF_LAST4 0 cglobal coeff_last4, 1,3 pxor mm2, mm2 LAST_MASK 4, r1d, r0 xor r1d, 0xff shr r1d, 4 BSR eax, r1d, 0x1f RET %endmacro INIT_MMX mmx2 COEFF_LAST4 INIT_MMX lzcnt COEFF_LAST4 %macro COEFF_LAST8 0 cglobal coeff_last8, 1,3 pxor m2, m2 LAST_MASK 8, r1d, r0 %if mmsize == 16 xor r1d, 0xffff shr r1d, 8 %else xor r1d, 0xff %endif BSR eax, r1d, 0x1f RET %endmacro %if ARCH_X86_64 == 0 INIT_MMX mmx2 COEFF_LAST8 %endif INIT_XMM sse2 COEFF_LAST8 INIT_XMM lzcnt COEFF_LAST8 %else ; !HIGH_BIT_DEPTH %macro LAST_MASK 3-4 %if %1 <= 8 movq mm0, [%3+ 0] %if %1 == 4 packsswb mm0, mm0 %else packsswb mm0, [%3+ 8] %endif pcmpeqb mm0, mm2 pmovmskb %2, mm0 %elif mmsize == 16 movdqa xmm0, [%3+ 0] packsswb xmm0, [%3+16] pcmpeqb xmm0, xmm2 pmovmskb %2, xmm0 %else movq mm0, [%3+ 0] movq mm1, [%3+16] packsswb mm0, [%3+ 8] packsswb mm1, [%3+24] pcmpeqb mm0, mm2 pcmpeqb mm1, mm2 pmovmskb %2, mm0 pmovmskb %4, mm1 shl %4, 8 or %2, %4 %endif %endmacro %macro COEFF_LAST48 0 %if ARCH_X86_64 cglobal coeff_last4, 1,1 BSR rax, [r0], 0x3f shr eax, 4 RET %else cglobal coeff_last4, 0,3 mov edx, r0mp mov eax, [edx+4] xor ecx, ecx test eax, eax cmovz eax, [edx] setnz cl BSR eax, eax, 0x1f shr eax, 4 lea eax, [eax+ecx*2] RET %endif cglobal coeff_last8, 1,3 pxor m2, m2 LAST_MASK 8, r1d, r0, r2d xor r1d, 0xff BSR eax, r1d, 0x1f RET %endmacro INIT_MMX mmx2 COEFF_LAST48 INIT_MMX lzcnt COEFF_LAST48 %endif ; HIGH_BIT_DEPTH %macro COEFF_LAST 0 cglobal coeff_last15, 1,3 pxor m2, m2 LAST_MASK 15, r1d, r0-SIZEOF_DCTCOEF, r2d xor r1d, 0xffff BSR eax, r1d, 0x1f dec eax RET cglobal coeff_last16, 1,3 pxor m2, m2 LAST_MASK 16, r1d, r0, r2d xor r1d, 0xffff BSR eax, r1d, 0x1f RET %if ARCH_X86_64 == 0 cglobal coeff_last64, 1, 4-mmsize/16 pxor m2, m2 LAST_MASK 16, r1d, r0+SIZEOF_DCTCOEF* 32, r3d LAST_MASK 16, r2d, r0+SIZEOF_DCTCOEF* 48, r3d shl r2d, 16 or r1d, r2d xor r1d, -1 jne .secondhalf LAST_MASK 16, r1d, r0+SIZEOF_DCTCOEF* 0, r3d LAST_MASK 16, r2d, r0+SIZEOF_DCTCOEF*16, r3d shl r2d, 16 or r1d, r2d not r1d BSR eax, r1d, 0x1f RET .secondhalf: BSR eax, r1d, 0x1f add eax, 32 RET %else cglobal coeff_last64, 1,3 pxor m2, m2 LAST_MASK 16, r1d, r0+SIZEOF_DCTCOEF* 0 LAST_MASK 16, r2d, r0+SIZEOF_DCTCOEF*16 shl r2d, 16 or r1d, r2d LAST_MASK 16, r2d, r0+SIZEOF_DCTCOEF*32 LAST_MASK 16, r0d, r0+SIZEOF_DCTCOEF*48 shl r0d, 16 or r2d, r0d shl r2, 32 or r1, r2 not r1 BSR rax, r1, 0x3f RET %endif %endmacro %if ARCH_X86_64 == 0 INIT_MMX mmx2 COEFF_LAST %endif INIT_XMM sse2 COEFF_LAST INIT_XMM lzcnt COEFF_LAST %macro LAST_MASK_AVX2 2 %if HIGH_BIT_DEPTH mova m0, [%2+ 0] packssdw m0, [%2+32] mova m1, [%2+64] packssdw m1, [%2+96] packsswb m0, m1 mova m1, [deinterleave_shufd] vpermd m0, m1, m0 %else mova m0, [%2+ 0] packsswb m0, [%2+32] vpermq m0, m0, q3120 %endif pcmpeqb m0, m2 pmovmskb %1, m0 %endmacro %if ARCH_X86_64 == 0 INIT_YMM avx2 cglobal coeff_last64, 1,2 pxor m2, m2 LAST_MASK_AVX2 r1d, r0+SIZEOF_DCTCOEF*32 xor r1d, -1 jne .secondhalf LAST_MASK_AVX2 r1d, r0+SIZEOF_DCTCOEF* 0 not r1d BSR eax, r1d, 0x1f RET .secondhalf: BSR eax, r1d, 0x1f add eax, 32 RET %else INIT_YMM avx2 cglobal coeff_last64, 1,3 pxor m2, m2 LAST_MASK_AVX2 r1d, r0+SIZEOF_DCTCOEF* 0 LAST_MASK_AVX2 r2d, r0+SIZEOF_DCTCOEF*32 shl r2, 32 or r1, r2 not r1 BSR rax, r1, 0x3f RET %endif %macro COEFF_LAST_AVX512 2 ; num, w/d cglobal coeff_last%1, 1,2 mova m0, [r0-(%1&1)*SIZEOF_DCTCOEF] vptestm%2 k0, m0, m0 %if %1 == 15 mov eax, 30 kmovw r1d, k0 lzcnt r1d, r1d sub eax, r1d %else kmovw eax, k0 lzcnt eax, eax xor eax, 31 %endif RET %endmacro %macro COEFF_LAST64_AVX512 1 ; w/d cglobal coeff_last64, 1,2 pxor xm0, xm0 vpcmp%1 k0, m0, [r0+0*64], 4 vpcmp%1 k1, m0, [r0+1*64], 4 %if HIGH_BIT_DEPTH vpcmp%1 k2, m0, [r0+2*64], 4 vpcmp%1 k3, m0, [r0+3*64], 4 kunpckwd k0, k1, k0 kunpckwd k1, k3, k2 %endif %if ARCH_X86_64 kunpckdq k0, k1, k0 kmovq rax, k0 lzcnt rax, rax xor eax, 63 %else kmovd r1d, k1 kmovd eax, k0 lzcnt r1d, r1d lzcnt eax, eax xor r1d, 32 cmovnz eax, r1d xor eax, 31 %endif RET %endmacro %if HIGH_BIT_DEPTH INIT_XMM avx512 COEFF_LAST_AVX512 4, d INIT_YMM avx512 COEFF_LAST_AVX512 8, d INIT_ZMM avx512 COEFF_LAST_AVX512 15, d COEFF_LAST_AVX512 16, d COEFF_LAST64_AVX512 d %else ; !HIGH_BIT_DEPTH INIT_XMM avx512 COEFF_LAST_AVX512 8, w INIT_YMM avx512 COEFF_LAST_AVX512 15, w COEFF_LAST_AVX512 16, w INIT_ZMM avx512 COEFF_LAST64_AVX512 w %endif ; !HIGH_BIT_DEPTH ;----------------------------------------------------------------------------- ; int coeff_level_run( dctcoef *dct, run_level_t *runlevel ) ;----------------------------------------------------------------------------- struc levelrun .last: resd 1 .mask: resd 1 align 16, resb 1 .level: resw 16 endstruc ; t6 = eax for return, t3 = ecx for shift, t[01] = r[01] for x86_64 args %if WIN64 DECLARE_REG_TMP 3,1,2,0,4,5,6 %elif ARCH_X86_64 DECLARE_REG_TMP 0,1,2,3,4,5,6 %else DECLARE_REG_TMP 6,3,2,1,4,5,0 %endif %macro COEFF_LEVELRUN 1 cglobal coeff_level_run%1,0,7 movifnidn t0, r0mp movifnidn t1, r1mp pxor m2, m2 xor t3d, t3d LAST_MASK %1, t5d, t0-(%1&1)*SIZEOF_DCTCOEF, t4d %if %1==15 shr t5d, 1 %elif %1==8 and t5d, 0xff %elif %1==4 and t5d, 0xf %endif xor t5d, (1<<%1)-1 mov [t1+levelrun.mask], t5d shl t5d, 32-%1 mov t4d, %1-1 LZCOUNT t3d, t5d, 0x1f xor t6d, t6d add t5d, t5d sub t4d, t3d shl t5d, t3b mov [t1+levelrun.last], t4d .loop: LZCOUNT t3d, t5d, 0x1f %if HIGH_BIT_DEPTH mov t2d, [t0+t4*4] %else mov t2w, [t0+t4*2] %endif inc t3d shl t5d, t3b %if HIGH_BIT_DEPTH mov [t1+t6*4+levelrun.level], t2d %else mov [t1+t6*2+levelrun.level], t2w %endif inc t6d sub t4d, t3d jge .loop RET %endmacro INIT_MMX mmx2 %if ARCH_X86_64 == 0 COEFF_LEVELRUN 15 COEFF_LEVELRUN 16 %endif COEFF_LEVELRUN 4 COEFF_LEVELRUN 8 INIT_XMM sse2 %if HIGH_BIT_DEPTH COEFF_LEVELRUN 8 %endif COEFF_LEVELRUN 15 COEFF_LEVELRUN 16 INIT_MMX lzcnt COEFF_LEVELRUN 4 %if HIGH_BIT_DEPTH == 0 COEFF_LEVELRUN 8 %endif INIT_XMM lzcnt %if HIGH_BIT_DEPTH COEFF_LEVELRUN 8 %endif COEFF_LEVELRUN 15 COEFF_LEVELRUN 16 ; Similar to the one above, but saves the DCT ; coefficients in m0/m1 so we don't have to load ; them later. %macro LAST_MASK_LUT 3 pxor xm5, xm5 %if %1 <= 8 mova m0, [%3] packsswb m2, m0, m0 %else mova xm0, [%3+ 0] mova xm1, [%3+16] packsswb xm2, xm0, xm1 %if mmsize==32 vinserti128 m0, m0, xm1, 1 %endif %endif pcmpeqb xm2, xm5 pmovmskb %2, xm2 %endmacro %macro COEFF_LEVELRUN_LUT 1 cglobal coeff_level_run%1,2,4+(%1/9) %if ARCH_X86_64 lea r5, [$$] %define GLOBAL +r5-$$ %else %define GLOBAL %endif LAST_MASK_LUT %1, eax, r0-(%1&1)*SIZEOF_DCTCOEF %if %1==15 shr eax, 1 %elif %1==8 and eax, 0xff %elif %1==4 and eax, 0xf %endif xor eax, (1<<%1)-1 mov [r1+levelrun.mask], eax %if %1==15 add eax, eax %endif %if %1 > 8 %if ARCH_X86_64 mov r4d, eax shr r4d, 8 %else movzx r4d, ah ; first 8 bits %endif %endif movzx r2d, al ; second 8 bits shl eax, 32-%1-(%1&1) LZCOUNT eax, eax, 0x1f mov r3d, %1-1 sub r3d, eax mov [r1+levelrun.last], r3d ; Here we abuse pshufb, combined with a lookup table, to do a gather ; operation based on a bitmask. For example: ; ; dct 15-8 (input): 0 0 4 0 0 -2 1 0 ; dct 7-0 (input): 0 0 -1 0 0 0 0 15 ; bitmask 1: 0 0 1 0 0 1 1 0 ; bitmask 2: 0 0 1 0 0 0 0 1 ; gather 15-8: 4 -2 1 __ __ __ __ __ ; gather 7-0: -1 15 __ __ __ __ __ __ ; levels (output): 4 -2 1 -1 15 __ __ __ __ __ __ __ __ __ __ __ ; ; The overlapping, dependent stores almost surely cause a mess of ; forwarding issues, but it's still enormously faster. %if %1 > 8 movzx eax, byte [popcnt_table+r4 GLOBAL] movzx r3d, byte [popcnt_table+r2 GLOBAL] %if mmsize==16 movh m3, [dct_coef_shuffle+r4*8 GLOBAL] movh m2, [dct_coef_shuffle+r2*8 GLOBAL] mova m4, [pw_256] ; Storing 8 bytes of shuffle constant and converting it (unpack + or) ; is neutral to slightly faster in local speed measurements, but it ; cuts the table size in half, which is surely a big cache win. punpcklbw m3, m3 punpcklbw m2, m2 por m3, m4 por m2, m4 pshufb m1, m3 pshufb m0, m2 mova [r1+levelrun.level], m1 ; This obnoxious unaligned store messes with store forwarding and ; stalls the CPU to no end, but merging the two registers before ; storing requires a variable 128-bit shift. Emulating this does ; work, but requires a lot of ops and the gain is tiny and ; inconsistent, so we'll err on the side of fewer instructions. movu [r1+rax*2+levelrun.level], m0 %else ; mmsize==32 movq xm2, [dct_coef_shuffle+r4*8 GLOBAL] vinserti128 m2, m2, [dct_coef_shuffle+r2*8 GLOBAL], 1 punpcklbw m2, m2 por m2, [pw_256] pshufb m0, m2 vextracti128 [r1+levelrun.level], m0, 1 movu [r1+rax*2+levelrun.level], xm0 %endif add eax, r3d %else movzx eax, byte [popcnt_table+r2 GLOBAL] movh m1, [dct_coef_shuffle+r2*8 GLOBAL] punpcklbw m1, m1 por m1, [pw_256] pshufb m0, m1 mova [r1+levelrun.level], m0 %endif RET %endmacro %if HIGH_BIT_DEPTH==0 INIT_MMX ssse3 COEFF_LEVELRUN_LUT 4 INIT_XMM ssse3 COEFF_LEVELRUN_LUT 8 COEFF_LEVELRUN_LUT 15 COEFF_LEVELRUN_LUT 16 INIT_MMX ssse3, lzcnt COEFF_LEVELRUN_LUT 4 INIT_XMM ssse3, lzcnt COEFF_LEVELRUN_LUT 8 COEFF_LEVELRUN_LUT 15 COEFF_LEVELRUN_LUT 16 INIT_XMM avx2 COEFF_LEVELRUN_LUT 15 COEFF_LEVELRUN_LUT 16 %endif